Abstract
Scientific visualization is well known to be useful in assisting user understanding of properties of particular data sets, especially those of scientific origin. Two important aspects of these data sets are (a) the data exists in some coordinate space and (b) one or more data values are associated with the coordinates. Typical applications include fluid dynamics and terrain data modeling. However, before scientific visualization techniques can be applied to programming, these two aspects must be clarified. Exactly, the two questions of interest are:

• How to define a structure consisting of a coordinate space for programming?
• How to define meaningful data values for such a coordinate space?

A new model, Geometric Representation of Programs (GRP) proposed by the speaker, defines a computation space as a group of computations (i.e., arbitrary statements of execution) that exist in a pre-defined Cartesian space. An iteration space is a specific (simple) example of a computation space that is defined by loop nest. A coordinate space is thus inherent in this representation. However, the computation spaces allowed in the GRP model are complicated by a number of factors. Given that visualization of programs in the GRP model is beneficial, it remains non-trivial to address the above questions.

This talk presents the GRP model in terms of computational spaces and then explores how scientific visualization can be used to better understand the nature of the represented programs. Preliminary results obtained from AVS Express, a high-end scientific visualization commercial software, will be presented.